Is Your Index Reader Really Atomic or Maybe Slow?
summary
Lucene 4.0 replaces the old IndexReader model with AtomicReader and CompositeReader and a segment-oriented DirectoryReader, makes all IndexReaders read-only, and adds reader contexts to map local to global doc IDs so segment searches run truly in parallel instead of doing expensive on‑the‑fly composite merges. Migration advice shows how to replace custom norms with DocValues + Similarity. Result: much faster queries and safer concurrency, but existing custom reader/codec code must be refactored and retested.
AI-written summary · about
chapters (9)
- 0:00Talk introduction
- 2:12Lucene index basics
- 8:29Composite index reader concept
- 12:40Performance issues with segments
- 19:57Index reader interface changes
- 24:21Atomic index reader support
- 27:22Directory and index reader updates
- 30:20Solution overview and summary
- 37:09Q&A and closing remarks
related talks
42mZero downtime index upgrade in Apache Solr
30mCaching for Speed: Practical Performance Optimization in OpenSearch & Lucene
40mWhat's evolving in Elasticsearch and Lucene?
37mAlgorithms & data-structures that power Lucene & ElasticSearch
25mApache Lucene: From Text Indexing to Artificial Intelligence
31mUnspoken overhead of the Inverted Index
40mLeveraging Lucene Skip Indexes for Faster Aggregations in OpenSearch
24mWritable Warm: Breaking the Read-Only Barrier in OpenSearch Tier...